Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
, « ä ARI Research Note 86-96
l <
o
INFORMATION STORAGE AND ACCESS IN DECISIONMAKING ORGANIZATIONS
Ghassan J. Bejjanl Morgan Stanley
and
Alexander H. Levis Massachusetts Institute of Technology
for
Contracting Officer's Representative George H. Lawrence
BASIC RESEARCH LABORATORY Milton R. Katz, Director
DTIC JAN 2 71987
n $ a
U. S. Army
Research Institute for the Behavioral and Social Sciences
November 1986
Approved for public rtltatt; distribution unlimited.
87 1 21 10
1 1
U. S. ARMY RESEARCH INSTITUTE
FOR THE BEHAVIORAL AND SOCIAL SCIENCES
A Field Operating Agency under the Jurisdiction of the
Deputy Chief of Staff for Personnel
WM. DARRYL HENDERSON
EDGAR M. JOHNSON JJ01" W J- Tcchnid Director Cwnn-nding
Research accomplished under contract for the Department of the Army
Massachusetts Institute of Technology
Technical review by
Steve Kronheim
T>MI rapen. M tubmimd bf tht contractor, hct bttn durtd fc rricne to Dtfimc Ttchnical Information Ctntcr (OTIC) to comply with ratyulatory roquircmentt. It It« bttn livtn no primary distribution etbtr than to OTIC •nd will bt available only iirewph OTIC or other raftrtnet lorvico» «uch m tht National Tachnieal Information Strvict (NTIt). Tht view ^ opinion*, «nd/or finding containtd in this report art thou of tht author («I and thouM not bt conttrutJ a< on officia' Otpaitmani of tht Army position, policy, or dtcition, unit» w dtiignatad by othtr official documa< /tation.
JL. UNCLASSIFIED
»ECUKITY CLASSIFICATION OP THIS PAOC (Whan Dm» Enfrtd)
REPORT DOCUMENTATION PAGE 1. RtPORT NUMICK
ARI Research Note 86-96 2. OOVT ACCESSION NO
4. TITLE (m* Suhiltl») INFORMATION STORAGE AND ACCESS IN DECISIONMAKING ORGANIZATIONS
». AuTHORfaJ
Ghassan J. Bajjanl and Alexander H. Levls
I. PKRPORMINO ORGANIZATION NAME AND ADDRESS
Laboratory for Information and Decision Systems Massachusetts Institute of Technology Cambridge, Massachusetts 02139
II. CONTROLLING OFFICE NAME AND ADDRESS
U.S. Army Research Institute for the Behavioral and Social Sciences, 5001 Eisenhower Avenue, Alexandria, Virginia 22333-5600
14. MONITORING AGENCY NAME • ADDRKSSf" SJBCSJ hum C—ttolUnt OUIe»)
READ INSTRUCTIONS BEFORE COMPLETING FORM
1. RECIPIENT'S CATALOG NUMBER
S. TYPE OF REPORT 4 PERIOD COVERED
Interim Report May 1985 4. PERFORMING ORO. REPORT NUMBER
4. CONTRACT OR GRANT NUMBERO)
MDA 903-83-C-0196
10. PROGRAM ELENENT. PROJECT, TASK AREA 4 WORK UNIT NUMBERS
2Q161102B74F
12. REPORT DATE
November 1986 IS. NUMBER OF PAGES
29 14. SECURITY CLASS, (el Ihlt rtport)
Unclassified IS«. DECLASSIFICATION/OOWNGRADINC
SCHEDULE
14. DISTRIBUTION STATEMENT (ol »Im Report;
Approved for public release; distribution unlimited.
! Acnrnsion Tor
IT. DISTRIBUTION STATEMENT (ol (ft* mbalraet mffd In Bleck 70. It dlllerenl Iren Hapert;
'TIS GI. iil OU UM
( Uuimnouno^d , Justlfioatlon
O D
14. SUPPLEMENTARY NOTES
George H. Lawrence, contracting officer's representative
This research partially funded by Office of Nava- Research
i Availability
IS. KEY WORDS (Ceollnue en reveree »Ide II neceeeery mtd Identity by ftlec* number;
Decisionmaking Database Organizational Structures Evaluation Organization Performance Workload
Dlst
A-(
Cedes Avail arli/vT
SpecieL
20. ABSTRACT (Cm**ue am referee et* H rmc—eetr md Identltr by Meek nunber;
Information storage and access in decisionmaking organizations is modelled us- ing a Petri Net representation. A centralized and a decentralized database con- figuration are analyzed, and their impacts on the decisionmakers' workload are assessed. Organizational protocols are defined, and their criteria of accept- ability presented. Protocols' key variables, minimum allowable input inter- arrival time, and response time are determined for two organizational structures - parallel and hierarchical. A numerical example suggests the use of timeliness
(over)
DO/, JAM 7» 1473 COfTlOM OF t NOV 4k It OBSOLETE UNCLASSIFIED
SECURITY CLASSIFICATION OF TMIf PAGE (Wlmt Dmlm entered)
J
UNCLASSIFIED «tfCUWTV CLAMiriCATIOM OT TMI« ^AOCO»»»«» «>•«• «»«•««
ARI Research Note 86-96
20 ABSTRACT (continued)
as a third organizational attribute - the first two being workload and performance. It also demonstrates the importance of updating coordination in evaluating the organization's performance.
UNCLASSIFIED^ 11 MCURITV CLASSiriCATtON Or TNM ^ACCrWhm Dim Bnffl)
mmma^ammmmma*mmm^mmm*^^^*m*M.ivii*M.mLiJki\jM\Jiiuk, _p.( _J»rj»r"_» rVM p^ '■■ * r^» n mm tau — n »h ia M n n rv*«».n.« r ^ftjiKJ^Kl^H^HMHH
INFORMATION STORAGE AND ACCESS IN DECISIONMAKING ORGANIZATIONS
by
Ghaasan J. BeJJani ••
Alexander H. Levia •••
\
ABSTRACT
Information storage and access In decislonmaklng organizations Is modelled using a Petrl Net representation. A centralized and a decen- tralized database configuration are analyzed, and their impacts on the declslonmakers' workload are assessed. Organizational protocols are defined, and their criteria of acceptability presented. Protocols1 key variables, minimum allowable input interarrival time, and response time are determined for two organizational structures: parallel and hierar- chical. A numerical example suggests the use of timeliness as a third organizational attribute - the first two being workload and performance. It also demonstrates the importance of updating coordination in eval- uating the organization's performance.
•This worlc was carried out at the KIT Laboratory for Information and Decision Systems with support provided In part by the Office of Naval Research under Contracts NOO014-83-K-O185 (NR 274-349) and N00014-84-K-0519 (NR 649-003) and In part by the OS Army Research Institute for the Behavioral and Social Sciences under Contract No. MDA903-83-C-0196.
•• Morgan Stanley, 35 Water Street. New York. NY 10041
••• Laboratory for Information and Decision Systems. MIT. Cambridge, MA 02139.
tw>^M^mmmmmmmmamamt\.-h 'A it L-W w uvv.vv% .v L-V WV Livukr-v LV v» L-W < t'JiL% VJW\IW\»AAI\*.L^> V\I. VNTJWV*
1. IMTRODOCnON
During th« past d«oad«, infm-aatlon theory has b*ea appllad to the
analysis and valuation of organisations. rlrst dovslopod by Shannon
(Shannon and Vaarar. 1949), information theory aaturad into a aathaaatioal
thaory in its own right, and was applied to the study of various
ooMunioations systeas (Gallager. 1968). It was then used ss a basic tool
for aodeling huaan deoisionasking (see Sheridan and Parrel, 1974, and
Drenick, 197S). The Partition Law of Znforaation (Conant, 1976) provided a
physical interpretation of the aathaaatioal expressions derived by using
the n-diaensionsl version of the theory.
A two-stage inforaation theoretic aodel of the deoisionsaker was
introduced by Boettcher and Levia (1912). Quantitative aeans for aeasuring
the huaan deoiaionaakers' workload and the organization's perforaanoe were
designed under a set of restrictive assuaptions. Subsequent research
effort (Hall and Levis, 1994} Chyen and Levis, 1985} Toaovlc and Levia,
1984) was oriented towards relaxing some of those assuaptions and resolving
acre ooaplex issues related to a realistic use of the deeisionaaking aodel.
This paper addresses the issue that deoiaionaakers are not aeaoryless
(an aasuaption in the original aodel) and that inforaation storage and
acccess devices are actually put to service in aost aodern organisationa.
The study of databaaes in aoyolioal organisationa la approohed along two
directions: (a) computation of modified activity terms that represent the
decisionaaker'a workload and (b) consideration of time and delays in the
normal functioning of an organization. The two directions are developed
aeparately but are brought together in the illustrative example of the
last section.
Pigure 1 shows the Petri Net representation (Tabak and Levis, 1984) of
the two-stage aodel of the ntb member of an organization. His input xn is
a component of a single vector source distributed by a set of partitioning
aatrioea aaong all the deoiaionaakers (Stabile and Levis, 1984). The
dedaionaaker processes this input in the situation sssessaent (SAn) stage
1\A.njA**UfcMBl*^UW\fc\«^WUW\Ä^^ 1A UV».VIT«Irt*U> UV V> w> .> V-HIJ
X -OH
L^
Figur« 1. Petri Net Representation of the nth Declslonmaker of the Organization
to cetei^lne or »elect a particular yalue of the rarlable z11 th»t denotes
the situation. Be may cooDunlcate his assessment of the situation to other
■embers (zn0) and be may receive their assessments In return (zon). This
supplementary Information may be used to modify his assessment. I.e.. It
may lead to a different value of zn denoted by zn'. Possible alternatives
of action are evaluated In the response selection (RSn) stage. The outcome
of this process Is the selection of a local action or decision y11 that may
be oonnunlcated to other team members or may form all or part of the
organization's response. A command Input from other dedslonmakers. v .
may affect the selection process.
The situation assessment stage consists of U algorithms (fj.
1-1.....0). The value taken by the variable un determines the algorithm to
ÄiamiüOüL>jflba*5U>jouo^jO'^^ ..•c.vfr^jvcnuwwAu, yiüi^VNiMj^wuAva
wmmmi^mwrwi i ■ ii ■■ w^n^mmmmrvwm
b« used, and is obosan according to the probability distribution p(un).
SiBilarly. the choice of an algorithm in the RS stage is determined by the
variable t11, with probability distribution pC^lz0').
As a response to the need for aemory and information handling In
today's organizations, the oonoept of Deoision Aids first appeared a 11 tie
■ore than a decade ago. These devices are evolving into well-integrated
Decision Support Systems (DSS) (Keen. 1981). The database is one of the
three main parts of a Decision Support System. The other two are an
information management program, and a machine-user interface (computer
terminal) (Sprague, 1980} Sprague and Carlson, 1982). This paper will
address the database and decisionmaker/machine interface Issues from an
Information theoretic point of view. The database's storage and access
procedures, and their Impact on the decisionmaker'a workload and
performance levels, will be described.
2. THE GENERAL DATABASE MODEL
The database model developed in this paper conforms to the traditional
definition of an information storage device: it can receive Information
from an external source, it stores it adequately, and it delivers this
information, or part of it. whenever accessed by its users. The Petri Net
model adopted here consists of two stages (see Fig. 2). The first stage,
transition C, receives an input from the decisionmaker who requests access
to the data. This input represents the situation in which the user is.
Transition C determines then the nature of the Information needed to cope
with that situation, and sends a query to the next stage, D. Transition D
performs the actual search, and delivers the data to the decisionmaker at a
predetermined stage of his internal decisionmaking process.
Figure 2. Petri Net Representation of the General Database Model
iv\>v^*vv\v^>:^v:.x>>:w>j.>u,%jMii^M^>jfl^^^
Databases oan be used In «Itbar a oeotrallzed or a daoentrallzed
configuration. Daoentrallzed databases are defined here as Individual
storage units, aooessed ezolusively by one declslonnaker. and holding and
deliTerlog inforaation relevant to this deolsionaaker'a task only. It was
proved (Bejjani, 198S) that the Increase in activity due to a centralized
or decentralized configuration were siallar. However, there are Important
differences. First, the time associated with the query process Is auch
shorter when the database is an individual one than when it is centralized.
In effect, in the former oase, no Irrelevant Information is to be scanned
and then discarded, which happens in the latter oase, and the system's
answer to its stimuli is more timely. However, an advantage of a
centralized database structure Is that It allows for more convenient
updating. It oan be updated in one operation, providing all the
dedsionmakers with equally recent information, whereas decentralized
databases require a much greater updating effort to obtain the same result.
This paper will develop information theoretic aspects of the centralized
databases and discuss the decentralized oase briefly (for a comprehensive
comparison of the two configurations, tbe reader is referred to BeJJanl,
198S).
2.1 Centralized Databases
A centralized database is a database shared by all members of the
organization. It is physically located in one place, and individual
terminals allow the dedsionmakers to access It independently. In the
Petri Net representation, a centralized database is modeled as one unit,
comprising several transition C/transition D sequences. There are two such
databases, one for the SA stege, called DBSA. and one for the HS stage.
DBRS. Tbe inputs to transition 0° in DBSA are the inputs to the arh
decisionmaker. x0, and the variable u11 indicating the SA algorithm he Is
about to use (see Fig. 3). Transition Cn emits then a message towards
transition 0° that carries a query for the information needed for DMn to
process z0 through the selected SA algorithm. Dn in turn delivers the
requested data, d|A. to the decisionmaker. who receives It as an input to
the algorithm he is using. The usage of DBRS follows a similar rationale
applied to the RS stage.
LlAMMhfiKQiK
Y l-o
Figure 3. Petri Net Representation of DM0 Using Centralized Databases
The use of databases has a significant impact on the decisionnaker's
workload, as can be seen In the following development. Activity rate terns
are derived by applying the Partition Law of Information Rates (Conant,
1976) to the deolslonmaklog model ured here. For a more complete
description of the calculations, the reader is referred to Bejjani (198S).
The modifications to the basic model are due to the presence of tno
supplementary variables. d|A and d§s, and to their relationship with the
existing structure. For simplicity, the superscript n will be omitted in
the following equations whenever confusion may not arise.
^AjOUOlVXVHWXMXVIULX^TUnLkKH.lUAMUMii ■ A* )UnhMB|M|aB|B|B|BMHHMiaHBI^HBaBjaHHBHaMH|BMHaHaHBHMHUl AjllUl AJMMMi
Throughput Hat«:
Ft - fU.d^.^.T^.d^u.z00.!.f.yno.y) ttl
Blookag« Rat«:
rb - i(i,^..<».T"',dlls) - rt <»
Itela« Hat«:
F - H(u) ♦ H_(T) (3) n T
Coordination Hat«;
0 Fc- } (|J(p(x.d3A))*^:H(pi))+H(z)
i-1 SA
*lt (p(z.zon)) *lj (p(I.y0n)) o o
T
S3
♦ H(z) ♦ H(2) + H(I.?) ♦ fx(x,dSA:zon) + Ts(x.dSi.zon:v0n)
* ^.»(x.^.«00."0"'-») ">
wh«r«
p1 - p(u-i) ; p. - p(v-J) (5)
MttftdiuuftflBttyaAffifi^^
■(p) ■ plog>p ♦ (l-p)log1(l-p) (6)
and oi la tha nuaber of Tariablaa of the algorlttaa 1 that ara reinitialized
at each iteration. The ayabol T^ designates the mean Intararrlval tiae of
the Input to the SI stage. TRS has an equivalent aeaning with respect to
the HS stage. The aean Input interarrival tiae can be used In the
equations» If the interarrival tiae la not constant, by regulating the
source (Hall, 1982). The functions |£. gjj"^, gj and jj are the individual
coordination rate functions of the Si, A, B, and RS algorithms, and are of
the following fora:
«i ■ 5 5u *ty - iu(wi) (7)
The terms H(z), H(Z), H(S,?) in (4) can be interpreted to represent
the direct coordination rate between aubsysteas, through the faot that one
subsystem's output is another's input. However, indirect coordination
between the subsystems is accounted for by the transmission rate terms.
f2(x,dSA:2on) represents the coordination rate that la due to the
relationship between x and dSA, and zon. Indeed, if tha Inputs to OH0 and
those to the rest of the organization (BO) are related, or If d|A and d^,
m J1 n, are not totally independent, due to the structure of the storage or
the updating process in the centralized database, then zon can bring to S
information about the inputs to the system that is not oootained in z.
Similar interpretations hold for the other two transmission rate terms.
The term T2,^(x,2on,dSA,von:dRS) raises the question of the relationship
between dSA and dHS, i.e., whether the situation assessment database (DBSA)
and the response selection one (DBBS) are related.
2.2 Decentralized Databases
A decentralized database structure is shown in Figure 4. The only
difference with respect to Figure 3 is the presence of only one transition
(^/transition Dn sequence per database, which aodels the exclusive use of
hÄ"^M>fr&^^
Y
HD
Figure 4. Petri Net Representation of DM0 Oaing Decentralized Databases
each database by a single declsionmaker. Apart from that, decentralized
databases are assuoed to function in exactly the same manner as centralized
ones.
2.3 Plzed Databases and the Memoryless Model
The results in section 2.1 were derived assuming the data dSA and d^
to be variable quantities. However, it might very well be the case that
dSA and d^ are fixed, either because the databases are never updated or
because the values taken by dSA and d^ remain valid during a very long
time, compared to the mean input interarrival time. In this simple case,
the database's direct contribution to the dcdsionmaker's activity rate is
null, and the expressions developed above become similar to those derived
in the basic memoryless declsionmaker case. They are derived by simply
eliminating the variables dSA and dj^ and the input variables to the
c^&^^mmma^
databases from the equations, whlob shows that the reduction fron the
database-equipped aodel to the aeaoryless one Is consistent.
3. PROTOCOLS AMD THEIR APPLICATION TO ORGANIZATIONAL STRUCTURES
3.1 Definition of Protocols and Determination of Their Key Tarlables
A protocol Is the description of the chronological order In which
elementary tasks have to be performed within one dedslonmaker as well
ss between two or more of them. Determination of procotols Is a
fundamental design problem for organizations In general* and of updatable
database-equipped ones In particular. Indeed, If the sequences of
operations for each dedslonmaker are not clearly defined, and If the
updating tempo of the database does not take these sequences Into account,
chaos can result. In brief, the situation could arise where different
declslonmakers would be accessing different databases at different times,
with different levels of accuracy and relevance of the data. In order to
process the same Input.
Since the Petrl Net representation (Tabak and Levls, 1984) clearly
Illustrates the organization's protocol as defined above and since a key
notion In the definition of a protocol Is the amount of time Involved at
each step of the dedslonmaklng process, an acceptable protocol for a given
organization will consist of Its Petrl Net representation supplemented with
the allocation of a processing time to each transition. The processing
time In fact represents the maximum allowable duration of a transition for
the orgadzatlon to function In an orderly fasdon, following Its operating
protocol.
Assumptions: In devising an acceptable protocol for the kind of
orgadzatlons dealt with here, the following assumptions are made:
(1) - the source emits the Input Z with a constant Interarrlval time
(2) - the various transitions have constant processing times.
(3) - ooHBunlcatlon between transitions Is Instantaneous.
10
^to&toM^^ «■fvr.vX ^^m^^SbÄ^M
(4) - any transition can process an Inoomlog input as soon as it has finished processing the previous on«* and no sooner.
(5) -no queusing is allowed at any stags of the process.
Assumptions (1) and (2) are a corollary of the broader assumptions that the whole systea operates in stsady state. Assumption (3) states in faot that all the dsoisionaakiog occurs within the transitions, and that no processing time is allooated to places. Assumption (4) is putting the "pipe-line effect* into words t this assures that the information flow through the system is continuous. Assumption (S) is a prerequisite to the application of Patri Met theory to the study of information theoretic deoisionmaking aodels: in effect, when queueing takes place, two or more different tokens can coexist in the same place. Since transitions do not have any means of recognizing priorities in choosing one token as an input out of the same place, the queue cannot be managed, and the organization's protocol is transgressed. (For a relaxation of this assumption, see Jin, 198S).
Proposition: Under assumptions (1) to (S), two necessary conditions for an organization's protocol to be acceptable are:
every transition in the system must have a processing time smaller than or equal to the mean Input interarrival time.
- the total amount of time spent by a token in one place cannot exceed the mean input interarrival time.
Both necessary conditions provide a symmetric analytical tool. Indeed, if the processing times of the transitions in the system are fixed, then the minimum admissible input interarrival time for the organization can be determined: It is equal to the greater of two quantities: the maximum processing time present on the Petri net diagram of that organization, and the maximum time any token spends in any place. Determining this minimum interarrival time is a very useful way of comparing the effectiveness of different organizational structures in a
given oontext.
The second necessary condition applies in oases of organizational
11
a>:^>v:::v:;;>:<>-
Interactions wb«re on« doolslonaaker sonda aoae information to anottaar and cannot proceed before receiving a message back. Thus, the proposition provides a way of determining the upper limit of tbe response time of this otber decisionmaker, everything else being fixed. This will be made clearer in tbe next section.
As a last comment, one sbould realize that tbe use of tbe proposition is not restricted to decisionmaking organizations. In fact, its argumeota are relevant to any acyclical information processing structure where Assumptions (1) to (5) are satisfied«
3.2 Construction of Protoools for tbe Centralized Case
In this section, tbe proposition will be used to develop protoools for two particular organizations using a centralized database configuration. Tbe basic quintity for each organization is f« tbe processing time of any SA or RS transition. It is assumed to be identical for all such transitions in both organizations, and it will be the unif used for all quantities computed here. Furthermore, T is assumed to be greater than the processing time of other types of transitions* on the grounds that more decisionmaking takes plaoe in SA and BS transitions than in the others. Tbe database's response time is assumed to be T as well.
Parallel Organizational Structure
In a parallel organizational structure, decisionmakers «re linked by somewhat symmetrical relationships: they do not formally Issue commands to each otber, and they can share information at all stages according to pre- established operating procedures. The parallel structure considered in this work is a three-person organization. (Fig. S) called "Organization P" from here on. DM1 and DM* use only one SA algorithm and two RS algorithms each, and DM* has tbe choice between two SA algorithms, whose output can be processed by only one RS algorithm. The command input von is absent from tbe model, due to the non-hierarchioal structure; the decisionmakers do however share information about their situation assessments.
12
a^Mam^^maMMiaMaMffl^^
Organization P uaes two oantrallzad databases. DBSA and DBRS; An
acceptable protocol for this organisation has been darlvsd and is given in
Figure S. Its sain characteristics are the ainlaua interarrlval time (IT)
it allows, T, and the organization's total response tine (RT), the tine
Interval between the arrival of the input and the generation of a
corresponding response* which is equal to 19T/3.
Figure S. Protocol of Organization P Using Centralized Databases
13
<iÄ&&^^
Hitrarohieal Organization Structur«
A hlararotaloal organisational struoture allows daclslonaakera to hava
an Influance on «ach othe."'s response selection. This Influence can be
represented by a command Input. von. The hierarchical structure analyzed
here Is a three-person organization, known as organization H, equipped with
centralized databases as shown In Figure 6.
Figure 6. Protocol of Organization H Using Centralized Databases
14
■^t^^^^^^
Organization H oonslsts of two declslonoakers who aotually contribute
to Its output. DM1 and ON*, and on« coordinating deolslonaaktr, DM*, who
analTiaa DM1'» and DM*'a situation assossaants in order to Issue a coonand
to than that carries bis instructions about the way the organization's
response should be constructed. DM* is not in contact with the
environaent, therefore he does not need an SA a tage, neither do DM* and DM*
need an information fusion transition. A. The three deoisionmakers In
organization H bays each two HS algorithms.
One acceptable protocol for organization H is that represented In
Figure 6. The minimum interarrival time. HT/3, is much greater than for
organization P. This is due to the relationship between transition f* and
ON*, where the information coming out of f* has to be processed by all
DMa's transitions before transition B* can be fired and the last token
leaves the place z1. Application of the syHstric argument of the
proposition's necessary conditions determined the mean interarrival time
as llt/3. The organization's response time is calculated quite simply in
this ess by adding all processing times along the path followed by the
original input and is ST. For more coaplez organizations, the System Array
approach la preferable for computing time delays (Tabak and Levls. 1984;
Jin. 1983).
3.3 Construction of Protocols for Decentralized Caae
It waa pointed out in section 2.2 that the only salient differences
between a centralized and a decentralized structure as defined here pertain
to transition D's processing time and the establishment of satisfactory
updating. In this section, transition D is assumed to require a total time
of T/3. which is half what was needed in the centralized configuration.
Again, this number depends greatly on the nature of the organization'a
decision support system.
Acceptable protocols for organizations P and H with decentralised
databases are given in Figures 7 and 8. respectively. The minlmun inter-
arrival time IT and the response time RT for each organization are T and
13
ti&mMim
12T/3 for the parallel one and 10r/3 and 7T for the hlerarohlcal one.
The reduction in the XT and IT. when oonpared to the centralized case», is due entirely to the shorter response tine of the database.
Ö
2T/3
Ö-
DBSA1
C« .0»
msi •SA
DBRS* C .0'
nä&tä*1**
hO LUo^
DM* u* M*
DBRS2
f<5
CrO-ä
T/3 ,T/S d|A
DBSA3 .
C» |0S
Y o 2T/3
W>^. DBRS3
Figure 7. Protocol of Organization P using Decentralized Databases
16
buiUMBUa*AtfAI(AAaAAA/<AtoWAnAaAUV^^ S \ . vi/<VLU WWW« ■WW)flU>J.>^->Jt»it wVU-k VVL-V wv. LVIVJ
Figur« 8. Protocol of Organization H using Dacentralizad Databases
3.4 Bsmarks
Each protocol in the previous sections has been derived under some
very specific conditions, in order to nice different organizations and
different database structures comparable along the same criteria. These
results are contingent upon using similar transition processing times for
both organizational structures.
17
^i^M^i/OMMÄÖÖ^^^
The mlnlJBua allowabl« laput lnt«rarrival tine la auota greater for a
hlerarohloal organization than for a parallel one. This followa from the
■ore coaplex aequenoes of tasks that bare to be performed In a hlerarohloal
organization before a new Input can be handled. The total response time Is
also greater for organization H than for organization ?, and the difference
Is due again to the Increased complexity.
The second observation is that, whatever the organization, a
decentralized database structure leada to Improved performance with respect
to time. In organization P, the decentralized structure leads to an 11%
Improvement in the response time over the corresponding centralized one.
while in organization H its leads to Improvements in both IT and HT of 9%
and 13% respectively. These results are due to the basic premise that
decentralized databases takes less time to perform the data query process
than centralized ones do. (The numerical results of the above two
paragraphs are summarized in Table 1).
TABLE 1. TIME CHARACTERISTICS OF ORGANIZATIONS P AND H
Centralized DB Decentralized DB
IT(P)
IT(H)
X
llT/3
X
10T/3
RT(P)
RT(H)
19T/3
8T
17T/3
7T
IT - Minimum Allowable Interarrlval Timet RT - Response Time
18
IflHHHAAAUHilMtoW:' ̂ -MMMM^tiMMMKbt^^
4. AM ILLÖSTHATIVE EXAMPLE
4.1 DMcrlption of thg Orgmnizatlons tegd
In this Motion, tactical organizations, one parallel «nd one
hierarchical, are used to address the issues that arise in the qualitative
evaluation of organizations, the problens raised by a lack of coordination
between several individual databases, and the trade-off between performance
and tineliness. (The ezaaple is developed in its entirety in BeJJani,
1983).
The first organization Is the parallel one (Organization P) in Figure
S. It consists of three naval battle groups defending a maritime front.
The first group, DM1, holds one extremity of the front, DM* holds the
center, and DM* the other end. The inputs received by the organization are
signals emitted by unidentified platforms (submarines, surface ships,
planes). The different deoisionmakers' tasks are to attempt to identify
the source of these signals (enemy or friends) in the SA stage, and to
select the appropriate response (fire, request identification, or take all
measures required to face an attack) in the RS stage.
The SA database provides information, obtained from intelligence
sources, that describes the codes the enemy could use when emitting the
kind of signals received by organization P. This Information will be
compared to the actual input to determine the letter's identity. The RS
database, DBRS, informs the deoisionmakers about the level of alert present
in their area at each iteration.
The second organization is the hierarchical one (Organization H) shown
in Figure 6. The context is the same as for organization P, but here only
DM* and DM* actually receive any external signals or select an active
response. DM* is a coordinator who. based upon the situation assessments
received from DM1 and DM*, gives instructions about what RS algorithm
should be selected by either of them. The organization's overall mission
is the one defined for organization P.
19
«MMMÄSa^^
4.2 BMttita
A prlaary f^atur« of the «xaapl« Is Ita nuatrloal slapllclty: all the
varlablas of the system are determined using binary logic based on the
comparison of quantities} there are no actual computations. Detailed
definition of the rarlables and the algorithms for the ease of a single
deolslonaaker has already been presented (Boettoher, 1981). The baslo step
In the computation of the performance-workload pair (J.G) Is determining
the pure strategies present In the organization (Levis and Boettoher,
1983). Zn the oases at hand, each IM has two pure strategies, each
obtained by the exclusive use of one algorithm (no deolslonsaker here has.
In any stage, more than two algorithms from which to choose). Activity
rates are a better measure of the deolslonaakers' workload than absolute
activity, because of the time constraints present In real-world situations.
The following equation applies here:
G - -2- 1 - 1,2.3 (8)
for either organization. The performance measure J is the expected value
of the cost the organization Incurs when it does not produce the correct
response for a given input. The workloads Q1 determined by each pure
strategy and the corresponding performance level J are plotted in the
(J, G1, Ga. Q1) space. Then, the performance-workload (P-W) locus for each
DM Is constructed where all possible mixed strategies are considered as
linear combinations of the pure ones. The graphs tuus obtained are the
projections of the overall (P-M) locus of the organization on each of three
planes: (GSJ). 0s,J), (G,,J). Because the Input is perfectly symmetric,
as well as DM1's and DM*'s roles in each organization, only the projections
for OH* and DM* are shown.
The use of activity rates in this Instance has the effect of
Illustrating very clearly the tradeoff between timeliness and workload
(Figs. 9 and 10). The performance of organization P is better than that of
B; the performance index for P takes values between 0 and 0.9 for P but
20
aBBifi^^^^
btttwMn 1.2 and 4.S for H. However, the vorkload of the oembers of P Is
■uch higher than that of H. naaely. It varies between t.l bits/sec and 11.6 for P, while It la only 1.3 and 2.7 for H. Measuring workload In terms of
aotlTlty rates gives organization H a slgnlfloant advantage as far as
keeping the deolslonoakers' workload below a given ■axlmun (the bounded
rationality oonstralnt) Is oonoemed. However, another tradeoff appears
here that Involves the notion of tlaellness. In effect, slnoe In this
example workload Is a decreasing function of the aean Interarrlval time,
Eq. (8), low workload levels are obtained by allowing a high IT, which
penalizes the organization In terms of Its timeliness. Thus, workload Is
reduced In H. but timeliness Is sacrificed.
Another consideration of Interest is the effect of poor updating
coordination on the organization*a performance when decentralized databases
are used (e.g., Figures 7 and I). The Impact of two different updating
sequences on performance Is reflected on the (P-W) loci. In the first
scheme. DM*'s and DM*'s RS databases are assumed to be updated, at T * 0,
In coordination with the Input arrival. DM1'a DBRS, however. Is updated
at x ••> T, with a delay of x over the Input to which the data correspond.
Mew performance levels for each pure organization strategy were derived
and a performance-workload locus was drawn (see Fig. 11(a)). The main
effoots are the upward movement of the original locus, and a degradation In
performance: the range of J Is from 0.3S to 1.0 as opposed to 0 to 0.9 for
the perfectly coordinated (or the centralized) database case; this
represents a drop of 29% In the average performance of the organization.
A second scheme exhibits a less coordinated updating sequence: DBRS*
Is updated at t + 0, DBBSX at T + v, and DBRS* at v •»> 2T. DM1 and DM*
both now have a greater propensity to make the wrong decision, and the
resulting (P-W) locus Is presented In Fig. 11(b). The best performance
(lowest J) Is now O.S, which Is very dose to what the worst performance
was In the coordinated case, and the worst performance level la 1.2. The
range of possible performance levels has shrunk further, and the drop In
average performance with respect to the original case Is 68%.
21
im*LmiU*VM.r^*.im*m*».^.m.^^^.~, 1||.| -,1 p, ^ ,-M^, .n „ „ njm^ij^j ! ■iiien<iikM.rjr>r^
Jf 4.9
4|-
3.9
3
2.5
2
t.5
lh
as
DM'
J I I L J I L 0 4 4.9 9 9.9 6 6.9 7 7.5 8 8.5 9 9.9 10 10.9 11 11.9 xi
Jt 4.5
4
3.5
3
2.5
2|-
4 4.9 9 9.9 6 8.9 7 7.9 8 8.5 9 9.5 10 10.5 it 11.5 »
Plgur« 9. (P-H) Loci for P Using Aotivlty Ha tos
22
V M JUdsfJSHIHAAMAJMm KMBlflLWMWkk > i/v uv \JU vii uu vn> L^ \ juvarukt a^i AA ui iu( ■«■^^■^■^■■■^■BUSMK'VJI v «vmi i
J1 I *-9 -
•^ ̂
4 - N 3.5 -
k V
3 - DM1
2.5 -
2 -
1.5 - L ̂
1
1 - i—•
as -
■ i i , , • i i
0" 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 5«
Jt 4.5
4
3.5
3h
2.5
2
1.5
I -
0.5 [•
o"
DM"
—' 1 1 1 1 1 1 I , 0.5 I 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 in
Plsur« 10. (P-V) Loci for H using Activity Bat«s
23
mmmmmmmmmmmmmmmimmmmmmmmammmmmmt*i*:**mL*jM&jto&<&tt^^
J4 4.5
4
3.5
3
2.5
2
1.5
as ' « « » I I I L J L. J I L
0*4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 10.5 II 11.5 ni
J 4.5
3.5
3
2.5
2
1.5
1 -
0.3
P- J ' « ■ ' ■ ' ■ J I ' ' 0 4 4.5 5 5.5 6 6.5 7 7.5 8 85 9 9.5 10 10.5 II 11.5-,
Figur« 11. (P-V) Loci for DM1 in P with uncoordinated DaUbases
m^K—^mi^a—^**M
L«t tbe transition prooMalng tin« b« «qunl to t oooonds. Then IT(P)
will b« t sooonda an vail, and ITCH) llx/3 aaoonda (or 10t/3 aao.(
depending on the database oonflguratlon). In any event. IT(H) la auch
greater than ZT(P) and nay handicap organization B If It has to respond to
threats arriving at a high rate. As an example, consider a wave of enemy
planes attacking the battlegroup: Zf no defense can be Initiated without
processing every Input through the SA stage, then the anti-aircraft
batteries cannot ahoot at a rate higher than one aissile every 3.6x sec.
Zf a nav threat arrives once every x aaoonda, then P is an adequate
structure, while H is far froa being one. An additional disadvantage of
organization B appears when response times are taken into account: the
battlegroup will need 8x aaoonda to fire on a threat after it la detected}
this might be too long if tbe threat is very close to the battlegroup when
its presence is detected.
When the platforms that the organization has to deal with are alower
units, like submarines or surface ships, organization B's timeliness
disadvantage is less critical, because of the longer time available for
constructing an adequate response and because of the smaller threat arrival
rate. In facw., the latter can be so small as to make any difference
between ZT(P) and ZT(B) seem irrelevant. Since the organizations designed
in this example have to deal with both alow and fast threats, one has to
consider the relative probability of being attacked by fast or slow
threats, and weigh it by the expected costs in each alternative during the
evaluation of the two organizational structures,
S. CONCLUSION
In this paper, the use of database networks was Introduced into the
organization, in two alternative configurations: centralized, and
decentralized. Information theoretic aspects of data storage devices were
analyzed. Time-related consideration were preaented and used to create new
criteria for the evaluation of the organization. The example illustrated
the major theoretical results.
25
**>***ÜMMMM^^ vCVttJttttl
$. REFEREMCES
Bajjanl* 0. J.. "Infomatlon Storage and Access in Declalon—king Organisations,* M.S. Thesis, LIDS-TH-1434, Laboratory for Inforaatlon and Decision Systeas, MIT. Cambridge, MA. January 198S.
Boettober. C. L. 'An Information Theoretic Model of the Decision Maker,* M.S. Thesis, LIDS-TB-1096, Laboratory for Information and Decision Systems. MIT. Cambridge. MA. June 1981.
Boettober. C. L. and A. H. Levis, 'Modeling the Interacting Decislonmaker with Bounded Rationality.* IEEE ITans. on Systems. Man, and Cybernetics. Vol. SMC-12. No. 3. May/June 1982.
Chyen. H. B-L. and A. B. Levis. •Analysis of Preprocessors and Decision Aids in Organisations.• to be presented st the 2nd IFAC/IFIP/IPORS/IEA Conference on Analysis. Design, and Evaluation of Man-Machine Systeas, Tsrese. Italy. September 10-12, 1985.
Conant, R. C, 'Laws of Information Hhloh Govern Systems.* IEEE Trans, on Systeas. Man, and Cybernetics. Vol. SMC-f. Bo. 4, April 1976.
Cothier. P. H., and A. B. Levis "Assessment of Timeliness in Cosmand and Control.* LIDS-P-1454, Laboratory for Information and Decision Systems. MIT. Cambridge. MA. April 198S.
Drenick. R. P., •Organization and Control,* Directions in Large Scale Systems. T.C. Bo and S. E. Hitter, Eds.. Plenum Press. New Icrk. 1976.
Gallager. R. G.. Information Theory and Bailable Communication. John Wiley and Sons. Inc., New York, 1968.
Ball. S. A.. "Information Theoretic Models of Storage and Memory.* MS Thesis, LIDS-TH-1232, Laboratory for Information and Decision Systems. MIT. Cambridge. MA. June 1982.
Ball. S. A., and A. B. Levis, •Information Theoretic Models of Memory in Human Deoisionmaking Models.* Preprints of the 9th Vorld Congress of IF AC, Vol. VI, Budapest. Hungary, July 2-6, 1984.
Jin. V. •Delays for Distributed Oecisionmaking Organizations.* M.S. Thesis. LIDS-TH-14S9, Laboratory for Information and Decision Systems, MIT, Cambridge, MA, May 1985.
Keen, P. G. V., •Value Analysis: Justifying Decision Support Systems,• MIS Qusrterly, March 1981.
Shannon, C. E. and W. Veaver, The Mathematical Theory of Ccmmunioation, Onivarsity of Illinois Press, 1949.
26
MÄiStfMiöiäUfli^^
1 • ft
Sheridan. T. B.. and W. I. Parrall. Man-Maohlna 3yat—a, KIT Praaa, 1974.
Spragua» Jr.« K. H., *A Pranevorlc for tba Davalopaant of DSS,* MI3 Quartarly. Oaoaabar 1980.
^>ragua. Jr.* B. B. and E. D. Carlson, Building Bffaotlya Daolalon Support Syatama. Prantioa-Ball. Englewood Cliffs, NJ. 1982.
Stabila, D. A. and A. B. Lavis, "The Oasign of Inforaation Structures: Basic Allocation Strategies for Organizations," Larga-Scala Syatama, Vol. 6. 1984. pp. 123-132.
TabaJc, D. and A. B. Levis, •Petri Bet Bepresentation of Decision Models,* Proc. 7th MIT/OBB Worlcshop on C* Syateaa, LIDS-B-1437. Laboratory for Inforaation and Decision Syatama. MIT. Cambridge. MA, December 1984.
Tbmovic. M. M.. and A. B. Levis, •On the Design of Organizational Structures for Command and Control,* Prop. 7th MIT/OHH Worlcshop on C* Syatama, LIDS-B-1437. Laboratory for Information and Decision Systems, MIT, Cambridge. MA. December 1984.
27
i«iaMö«iöjaÄöötQö^>i